AI’s capabilities in cybersecurity have just taken another major step forward.
OpenAI has officially released GPT-6 Cyber, its new-generation cybersecurity model. It can autonomously discover software vulnerabilities, verify whether they are exploitable, and help researchers develop fixes.
Compared with the previous generation, GPT-5.6 Cyber, the new model improves its exploit-testing results by about 51% in relative terms, and it follows instructions much better when carrying out security tasks. More notably, OpenAI has already put GPT-6 Cyber to work on real software security research, with striking results.
In its Linux kernel research, GPT-6 Cyber successfully developed 948 working exploits. In Firefox, it found 81 previously unknown high-severity vulnerabilities. Some of the Firefox vulnerabilities have already been fixed by Mozilla, and the security updates are rolling out to users.
